Ultrasound guided waves are sensitive to the biomechanical properties of long bones, and have been widely used in the health evaluation of long bones. Due to the high attenuation characteristics of bones, accurate estimation of the dispersion pattern of broadband coherent guided wave signals in long bones, and then accurate inversion of the material characteristic parameters of long bones, is a research hotspot and difficulty in the field of bone ultrasound. In this paper, a guided wave number estimation method based on the subspace of broadband coherent signal is proposed to extract the guided wave dispersion pattern with high accuracy; on this basis, an ant colony algorithm with global optimization is applied to estimate the material parameters of the long bone from the guided wave pattern, and the quantitative evaluation of the condition of the long bone is successfully realized. Simulation and ex vivo experiments (2 simulated samples and 2 ex vivo bovine bones) jointly verified the validity of the method, and the relative errors between the experimental dispersion and the reference dispersion were 3. 52% , 3. 83% , 3. 35% , and 4. 51% , respectively; for the ex vivo bovine bones, the average relative errors between the estimated values of the cortical bone thickness, longitudinal wave velocity, and transverse wave velocity and the reference values were 3. 10% , 0. 11% , and 0. 03% . In conclusion, the method proposed in this paper provides a new reference for quantitative ultrasonic long bone evaluation, and can also be applied to other solid waveguide dispersion extraction and structural health characterization with high accuracy.